Services like Azure Machine Learning and Amazon Machine Learning are publicly available alternatives that provide similar end-to-end platform functionality but only integrate with other Amazon or Microsoft services for the data storage and deployment components of the pipeline.
Key goals of a REST-based platform include scalability of components interactions, generality of interfaces, independent deployment of components and intermediary components to enforce security and reduce latency.
This model only includes deployment and component aspects of the system, showing the software elements that conforms it: platform executable, OpenCV libraries, the stages that conforms the pipeline, and the components that can be assembled in each stage.
A key decision towards algorithm implementation was the adoption of the REST architectural style, because it is suitable for achieving three important goals: independent deployment of components, ease of standardised communication between components and generality of interfaces.
These changes in the deployment of components are represented by the current network of components of all applications and the new network of components.
Let (x = left( {{x_{cs}}} right) ) ((c in tilde{C}), (s in S)) be a candidate solution of deployment of new components where component c is deployed on server s if ({x_{cs}} = 1), or not if ({x_{cs}} = 0).
However, the problem of interpreting results from quantitative analysis of extra-functional properties is still challenging because it is hard to understand how the analysis results (e.g., response time, data confidentiality, mean time to failure) trace back to the architectural model elements (i.e., software components, interactions among components, deployment nodes).
Upon deployment, a LooCI component registers with the local Reconfiguration Engine, which supports introspection of component state and life cycle control.
Participants additionally highlighted the need for direct model deployment and execution (component binaries, execution scripts, etc).
One main reason is the lack of properties favoring incremental deployment, an essential component for new technology adoption.
Paraiso et al. [11] present a so-called federated multi-cloud PaaS infrastructure that enables the deployment of service component architecture (SCA) applications on heterogeneous PaaS and IaaS offerings.
